110 likes | 278 Views
Mapping for Reuse in Knowledge-Based Systems. DERI Research Seminar – Ontology Mediation Track 2004-05-28 Jos de Bruijn Digital Enterprise Research Institute jos.de-bruijn@deri.ie. References for this presentation.
E N D
Mapping for Reuse in Knowledge-Based Systems DERI Research Seminar – Ontology Mediation Track 2004-05-28 Jos de Bruijn Digital Enterprise Research Institute jos.de-bruijn@deri.ie
References for this presentation • J. Y. Park, J. H. Gennari, and M. A. Musen (1998). Mappings for reuse in knowledge-based systems. In Proceedings of the 11th Workshop on Knowledge Acquisition, Modelling and Management (KAW 98), Banff, Canada. Jos de Bruijn <jos.de-bruijn@deri.ie>
Agenda • Reusable Components in KBS • Mappings between KBS components • Implicit Mappings • Procedural Mappings • Declarative Mappings • A Mapping Ontology • Requirements • Mapping Properties • Structure of the Ontology • Mapping Patterns • Conclusions Jos de Bruijn <jos.de-bruijn@deri.ie>
Reusable Components • Knowledge-Based Systems consist of: • Domain Knowledge • Domain Ontology • Knowledge Base instances • Domain-independent Problem-Solving Methods (PSM) • Input and output described by method ontology • For a given task, relevant domain knowledge and PSM may already exist • In order for a PSM to solve a domain-specific task: • PSM needs to be connected to domain knowledge • Mapping Relations define translation of concepts and objects Jos de Bruijn <jos.de-bruijn@deri.ie>
Mappings between KBS components • Mappings are defined as: • Whatever mechanisms are used to convert between structures existing in one component and analogous structures expected by another • Types of mappings: • Implicit mapping: • also component adaptation. • The object definitions in one component are changed to fit the requirements of the other • Procedural mappings: • translation code to convert instances from one representation to another. • Difficult to reuse • Declarative mappings: • Easier to reuse • Mappings convey the intent of the mapping: what instead of how Jos de Bruijn <jos.de-bruijn@deri.ie>
Mapping Ontology - Requirements • Expressiveness • Ease of Use • Design of mappings should be straightforward • Clarity • Mappings should be easy to reuse and comprehend • Parsimony • Set of mapping relations should be minimized • Number of classes in ontology should be kept small • Efficiency • Principled design/natural distinctions • Should be based on both theoretical and practical requirements of mapping task Jos de Bruijn <jos.de-bruijn@deri.ie>
Mapping Ontology – Mapping the properties • Complex objects are decomposed into simple objects (corresponding to properties) for mapping • Four dimensions for the mapping task: • Power/complexity • Scope • The range of domain classes • Dynamicity • When and how a mapping should be invoked (precompiled vs. run-time) • Cardinality • 1-1, 1-n, n-1 Jos de Bruijn <jos.de-bruijn@deri.ie>
Mapping Ontology – Mapping the properties Jos de Bruijn <jos.de-bruijn@deri.ie>
Mapping Ontology – Structure of the Ontology Jos de Bruijn <jos.de-bruijn@deri.ie>
Mapping Patterns • Mapping patterns capture sets of mapping relations applied to similar domain-to-method conversions • Benefits: • Less repetition in mapping specification • Increased reliability • Reduced verbosity of the mappings • Convey the intent of the mapping Jos de Bruijn <jos.de-bruijn@deri.ie>
Questions? Jos de Bruijn <jos.de-bruijn@deri.ie>